Redistribution between multiple routing protocols

[sheerpace]

Hi,

One of our challenges with the new upgrade project is to use controlled redistribution between multiple routing protocols.

The network is predominantly a VSAT based network covering 20 sites. Few of those sites have have MPLS connectivity. Service provider cloud runs BGP on their backbone. Our preferred routing protocol within our organization is OSPF. But a product used in our VSAT network supports only RIPv2.

There may be an instance where we may need to do a redistribution from OSPF into RIP into BGP into OSPF for an onward path alone. We are still in design stage, and there are many challenges and counter-challenges with the redistribution and post deployment support.

LAN-OSPF-OSPF/RIP-RIP-RIP-RIP/BGP-BGP-BGP/OSPF-SERVER Farm (Onward path)

SERVER FARM-OSPF-OSPF/BGP-BGP-BGP/RIP-RIP-RIP-RIP/OSPF-OSPF-LAN (Return path)

Kindly suggest if the timers are tuned to include satellite delay, can we go ahead with above proposal.

Thanks.

Rgds, Vj.

[jstoecker]

So, are you saying your connectivity for any given site of the twenty is

server farm - connected via OSPF router to

network/vlan - connected via OSPF router to

network/vlan - connected via BGP router to

service provider - connected via BGP router to

service provider - connected via BGP router to

network/vlan - connected via VSAT RIPv2 router to

network/vlan - connected via VSAT RIPv2 router to

network/vlan - connected via VSAT RIPv2 router to

network/vlan - connected via VSAT RIPv2 router to

network/vlan - connected via OSPF router to

network/vlan - connected via OSPF router to

workstation lan?

Out of curiosity, how much bandwidth is available on the VSAT? My experience with VSAT was with low bandwidth/hundreds of shared up/downlinks on the same subnet, and just one hop. At 9600 baud per node, there was no 'room' for a dynamic routing protocol when running IP [smile]. IP worked with the multi-second propagation delay for that one hop.

John

[sheerpace]

John,

Thanks for the reply.

We are providing bandwidth to the remote sites with no less than 128kbps and can move upto 1-2 Mbps. Yes, propagation delay cannot be changed.

To answer first part of your question,am bit confused.

1. Server farm directly connected to a L2/L3 Cisco Core LAN Switch.

2. This Switch connects to 2 Cisco Routers running HSRP b/w them. These Cisco routers run OSPF process for the LAN behind them.

3. 2 Cisco Routers have got 2nd Ethernet port which connects to another LAN switch which also hosts multiple satellite modems (which can talk only RIPv2).

4. 2 Cisco Routers and Satellite modems share Network knowledge using common dynamic routing protocol RIPv2. Hence, we need to redistribute OSPF into RIP and vice-versa.

5. Satellite modem (/hub) at one end talks to other satellite modem (/hub) at remote end using RIPv2. It then connects to a PE router in the Service provider's MPLS cloud. Means, we need to redistribute RIP into BGP and vice-versa.

6. To connect one of our Data centres (running OSPF)to this VSAT network, we need to redistribute BGP into OSPF and vice-versa.

In short, a packet from a client in one of the remote sites, takes a tedious journey, starting with LAN -> OSPF -> OSPF/RIP redistribution -> RIP<>RIP -> RIP/BGP -> BGP/OSPF redistribution -> Server farm ->OSPF/BGP redistribution -> BGP/RIP -> RIP<>RIP -> RIP/OSPF redistribution -> (OSPF) LAN

You're almost there. Hope this explanation throws more light on complex routing design.

Rgds, VJ.

[jstoecker]

VJ,

I'm still a little unclear on parts of the connectivity.

I've drawn a diagram that I think corresponds to what you're describing (mostly no routing protocols, only devices and segments) - see the attached. Let's get the diagram right, and then we can more intelligently talk about the individual pieces that make up the puzzle.

John

[sheerpace]

John,

Yes, you have drawn the correct diagram. I will also try to construct one sooner.

Thnks, VJ

[jstoecker]

OK, now to describe the routing protocols present in the design, and gather more info - correct me if this is wrong:

Routers R1 and R2 are running OSPF on their Switch 1 interfaces, and RIPv2 on their switch 2 interfaces.

The R3 router is running RIPv2 on the switch 3 interface, and eBGP on the MPLS interface.

The R4 router is running eBGP on the MPLS interface, and OSPF on segment four (that switch should be switch 4 not switch 3).

The idea behind this design is that segment 1 is reachable from segment 4/the OSPF data center area and vice versa, using dynamic routing protocols so that individual static routing entries are not needed.

Is segment 1 considered a remote site, equivalent to the other 19?

Does segment 1 require reachability to the other remote sites or just to the data center?

Can the satellite modems be configured with static routes?

John

[sheerpace]

John,

Excellent. You have got it correct reg Routing protocols. Yes, segment 1 should be able to interact with data center.

Considering segment 1 as a hub site, and has got 6-8 spokes (out of total 19); spokes should be able to reach both segment 1 and data center.

In our network we have identified 3 hub sites.

HUB A (Satellite Teleport): Which has got connectivity to all 19 sites (spokes).

HUB B (Region 1): 6-8 spokes, these spokes need to talk to HUB B & to HUB A.

HUB C (Region 2): 6-8 spokes, these spokes need to talk to HUB C & to HUB A.

HUB A & HUB B has got connectivity to MPLS cloud.

Traffic Flow (Primary Path).

HUB B to Data center: MPLS

HUB B to its Spokes: VSAT

(Backup path)

HUB B to Data center: VSAT to HUB A + MPLS to DC

HUB B to its Spokes: MPLS to HUB A+ (HUB A) vsat (Spokes)

HUB C to Data center: VSAT to HUB A + MPLS to DC

HUB C to its spokes: VSAT

HTH.

Rgds, VJ.